Identification de sources acoustiques au passage d'un véhicule routier par imagerie acoustique parcimonieuse dans le domaine temporel

Abstract : The study detailed in this manuscript is part of the effort to characterize the noise emission from road vehicles. We wish to identify the noise sources of a moving vehicle, when driven on a roadway in real-world conditions, with roadside acoustic measurements. The current acoustic imaging methods do not provide sufficient performance on road vehicles. A state of the art led to the selection of an existing method, MSA-PSF, which consists in deconvolving signals from mobile sources in the frequency domain under certain assumptions, and was originally developed for aeroacoustics. This method is adapted here to the context of road vehicles. Then, an original approach is proposed in order to tackle the specific constraints of this context: CLEANT. This is an iterative method, performed in time domain with a wideband approach, which takes into account the effect of sources motion and includes two parameters designed to refine the result: the loop factor and the stopping criterium. A further version of the algorithm, including a frequency filter, is also proposed and shows significant improvement in identifying secondary sources in some particular cases. An interesting point of CLEANT is the availability of the sources reconstructed time signals, which enables other types of analysis, especially the use of the coherence with signals from on-board measurements in order to separate the contributions of uncorrelated sources. MSA-PSF and CLEANT are evaluated with numerical simulations and a set of indicators to measure their source localization and quantification performance. They are then tested in a controlled laboratory conditions experiment, using a moving source. This experiment represents a first application of the methods to a practical case, involving a linear motion, two simultaneous sources and different kinds of signals (tone and wideband). They are finally compared to the classical approach of moving source beamforming, within the frame of an experiment on a road vehicle, in real-world conditions. The original approach CLEANT yields very encouraging results, and is a clear improvement from the conventional beamforming, especially at low frequency for the tested cases. Applying it to a road vehicle in real-world conditions highlights a potentially troublesome behavior of the method, and the solution brought by CLEANT's frequency filtered version, or by adapting its various parameters. The coherence with reference signals to discriminate the physical origins of the sources is also tested and underlines the role of the short duration of the signals related to the sources passing-by context.
Complete list of metadatas
Contributor : Abes Star <>
Submitted on : Wednesday, July 17, 2019 - 5:13:09 PM
Last modification on : Thursday, July 18, 2019 - 1:29:29 AM


Files produced by the author(s)


  • HAL Id : tel-02078608, version 2


Rémi Cousson. Identification de sources acoustiques au passage d'un véhicule routier par imagerie acoustique parcimonieuse dans le domaine temporel. Acoustique [physics.class-ph]. Université de Lyon, 2018. Français. ⟨NNT : 2018LYSEI106⟩. ⟨tel-02078608v2⟩



Record views


Files downloads